As SOA deployments become more important to running a business, reliability, scalability, availability and affordability will be key issues -- issues that grid computing concepts may help solve.
Vendors are taking different tacks in this area, but the intent is to "move the concept of grid into the world of business applications," said Massimo Pezzini, vice president and distinguished analyst at Gartner Inc. "Mainstream enterprises are trying to make sense of where to deploy SOA for business-critical applications to support demanding and extreme requirements from a reliability/scalabililty [standpoint]."
According to Jamie Bernardin, chief technology officer and co-founder of DataSynapse Inc., based in New York City, "Grid is a great platform for SOA to be deployed on. When you think about the kind of decoupling and shareablity you get out of the SOA model, it maps well onto a grid -- a shared usage platform. You can put policy and control around sharing, and scale out heavily used services. An adaptive infrastructure can satisfy business and technical demands."
DataSynapse recently announced FabricServer, infrastructure software that virtualizes transactional applications, enabling organizations to scale and share application server environments. FabricServer, which supports Java and .NET environments, allocates system resources based on business priority at runtime.
FabricServer "virtualizes Web applications, as well as Web services that run inside traditional Java application servers," Bernardin explained. "[It] manages deployment, activation [and] adaptive load balancing across application servers. So you have a grid running WebLogic, Tomcat, JBoss, Oracle. You can deploy to any container you want. FabricServer allows you to seamlessly move from one container to another; you could describe it as a distributed container of containers."
Bernardin said FabricServer enables the consolidation and commoditization of J2EE. "Large institutions have hundreds of Web applications and they're moving to Web services. There is the complexity of static provisioning those applications. And managing them separately from a runtime and operations standpoint is costly. If [organizations] can unify deployment, management and operations of all heterogeneous applications and application servers they use in one fabric, they've got cost savings of people and management and ease of operations. You're also looking at making better use of hardware resources [by not] overprovisioning application installations.
The benefit, he said, is that "businesses can put power as needed behind applications that are more important. They have power on demand to satisfy demands as they increase through the day or through planned usage. They can add generic compute nodes on the grid without rearchitecting."
While DataSynapse virtualizes applications across application servers, the recently announced Enterprise Application Fabric 3.0 from Appistry Inc., St. Louis, Mo., virtualizes applications across commodity hardware. The Appistry Enterprise Application Fabric aims to create reliable, stateful environments for Web services running on commodity-grade infrastructure, while providing application-level fault tolerance and scalability.
This creates a better ecosystem in which to deploy SOA, said Sam Charrington, Appistry vice president of product management and marketing. "The top line you're trying to achieve by adopting SOA is agility; SOA is great about providing agility from the perspective of software infrastructure," he said. "What SOA doesn't help with is agility as it pertains to physical infrastructure. If you have a service that is mission-critical, you end up being encumbered by a traditional IT approach."
With a traditional IT approach, if a service really takes off, Charrington said, organizations may think they have to put it on an expensive box to meet availability and performance demands. "But this is not the agility of SOA; it takes away from the huge benefit of SOA."
This is where virtualization comes into play, he said. "By virtualizing applications across a commodity infrastructure, managed by, or as part of an application fabric, you're able to get the full benefit that SOA offers."
Applistry Enterprise Application Fabric runs on top of the operating system, Charrington said. "You might use it in conjunction with an application server, but many customers are using Enterprise Application Fabric in the same way they'd use an application server," he said.
Appistry is leveraging grid concepts by "splitting/deploying applications on top of interconnected small boxes," Pezzini said. He said other vendors with similar offerings include London-based Paremus and Majitek in San Francisco. "They're trying to implement what is sometimes called the enterprise or transactional grid," he said.
An enterprise grid is one of many alternatives to traditional computing architecture, Pezzini said. "Companies are adopting SOA for business-critical scenarios. They also have a problem of cost, and they're looking for ways to reduce the cost of deploying SOA-type applications. If those products [like those from Appistry] prove to work for real, it might be an alternative platform to J2EE or .NET or the mainframe." However, he said, the vendors are young, as is the technology.
"It's unproven technology, clearly for leading-edge enterprises," Pezzini said. "But I definitely see it as one of the technology elements that will compose the architecture for the future."